CN103036792A - Transmitting and scheduling method for maximizing minimal equity multiple data streams - Google Patents

Abstract

The invention provides a transmitting and scheduling method for maximizing minimal equity multiple data streams, and belongs to the field of network data flow optimization in a computer network. The method comprises the following steps: inputting data center network topology information and data stream transmission request information and calculating an intermediary characteristic value of each edge in a network topology, wherein the data center network topology information contains a link connection relationship among all data centers and a bandwidth capacity of each link, and the data stream transmission request information contains a transmitting end and a destination end of each data stream transmission request as well as amount of data requested to be transmitted; evaluating different paths between two points based on the intermediary characteristic value, selecting a specific set Pi of K non-superimposable transmission paths for each data stream transmission request; and iteratively obtaining an optimal network bandwidth resource allocation scheme capable of satisfying the rules of maximizing a minimal equity corresponding to the specific set Pi.

Description

The minimum fair multiple data stream transmission dispatching method of a kind of maximization

Technical field

The network traffic data that the invention belongs in the computer network is optimized the field, is specifically related to the minimum fair multiple data stream transmission dispatching method of a kind of maximization, is used for solving selection of transmission paths and allocated bandwidth problem between a plurality of data flow of network.

Background technology

The route of data flow and bandwidth resource allocation problem thereof are one of the basic research problems in network traffics optimization field.When especially considering towards a plurality of different request competition shared network bandwidth resources, this problem more complex.On the one hand, need to distribute bandwidth resources as much as possible to improve network resource utilization for each request; On the other hand, need the fairness between each request of assurance, seize the situation generation that too much common network resource causes another part request " to be died of hunger " to prevent the part request.

In order between the resource utilization of network integral body and bandwidth on demand resource distributional equity, to obtain compromise, the flow network field is extensively adopted " the minimum fair rule of maximization (Max-Min Fairness; MMF) " that Internet resources are distributed and (be please refer to D.Bertsekas and R.Gallager, Data Networks, 2 ^NdEd., Englewood Cliffs, NJ:Prentice-Hall, 1992).So-called " the minimum fair rule of maximization " refers to: divide timing network bandwidth resources is carried out in a plurality of data flow requests, the request resource that no thoroughfare seizes existing stock number request still less increases the own resource of obtaining arbitrarily.

Existing correlative study to " the minimum fair rule of maximization " mainly concentrate on data flow transmission route fixing situation (please refer to (1) Y.Afek, Y.Mansour, and Z.Ostfeld, " Convergence complexity of optimistic rate based flow control algorithms; " in Proceedings of the twenty-eighth annual ACM symposium on Theory of computing.ACM, 1996, pp.89-98; (2) Y.Afek, Y.Mansour, and Z.Ostfeld, " Phantom:A simple and effective flow control scheme, " ACM SIGCOMM Computer Communication Review, vol.26, no.4, pp.169-182,1996; (3) Y.Bartal, M.Farach-Colton, S.Yooseph, and L.Zhang, " Fast; fair; and frugal bandwidth allocation in atm networks; " in Proceedings of the tenth annual ACM-SIAM symposium on Discrete algorithms.Society for Industrial and Applied Mathematics, 1999, pp.92-101), and the scene variable to transmission path, such as the MPLS network environment, these achievements in research often can't directly be used.On the other hand, have now for the minimum fair many Commodity Flows route of the maximization under the variable scene of transmission path and allocated bandwidth optimal scheduling algorithm and (please refer to (1) M.Allalouf and Y.Shavitt, " Centralized and distributed algorithms for routing and weighted max-min fair bandwidth allocation; " Networking, IEEE/ACM Transactions on, vol.16, no.5, pp.1015-1024,2008 and (2) D.Nace, L.Nhat Doan, O.Klopfenstein, and A.Bashllari, " Max-min fairness in multi-commodity flows, " Computers ﹠amp; Operations Research, vol.35, no.2, pp.557-573,2008), adopt simple exhaustive routing policy to solve whole path candidates, and obtain on this basis the allocated bandwidth solution of global optimum, but this can cause the computing cost of algorithm large, and autgmentability is poor.Especially when network topology scale and request increase in size, owing to the overlong time of finding the solution of algorithm, make it be difficult to be applied to actual network traffics and optimize in the scene.

Summary of the invention

The object of the invention is to solve a difficult problem that exists in the above-mentioned prior art, provide a kind of maximization minimum fair multiple data stream transmission dispatching method, by choosing specific set of paths for each request, and then reduce the mode of problem solving scale, reduce the time overhead of Algorithm for Solving, that is to say in the face of a plurality of data flow request the time, guaranteeing under the minimum fair regular prerequisite of maximization, maximization network bandwidth resources utilance, thereby reduce the time overhead of finding the solution of algorithm, make it can be applied to network size and the larger network traffics optimized transmission scheduling scene of request.

The present invention is achieved by the following technical solutions: the minimum fair multiple data stream transmission dispatching method of a kind of maximization comprises:

Input data center network topology information and data stream transmitting solicited message, the intermediary characteristic value on every limit in the computing network topology; Described data center network topology information comprises that the link connection between each data center concerns and the bandwidth capacity of every link; Described data stream transmitting solicited message comprises the data volume of transmitting terminal, destination and the request transmission of each data stream transmitting request;

Based on described intermediary characteristic value, the different paths of point-to-point transmission are assessed, for the not set P in overlapping transmission path of specific K bar is selected in each data stream transmitting request _iAnd

Based on the described K bar of each the data stream transmitting request set P in overlapping transmission path not _i, iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision that maximizes minimum fair rule.

Described based on described intermediary characteristic value, the different paths of point-to-point transmission are assessed, for the not set P in overlapping transmission path of specific K bar is selected in each data stream transmitting request _iComprise:

(A1): given network G (V, E), source point and point of destination are respectively v _sAnd v _t, and the path number K that need to obtain;

(A2): find the solution its shortest path p1 in network G, p1 is added set Ps, and the bandwidth resources that p1 is occupied deduct from G, obtain remaining network G r; Described Ps is the set of paths of distributing to each data stream transmitting request;

(A3): repeating step on remaining network G r (A2) maybe can't be obtained new shortest path until number of path reaches K; If the situation in the path of many equal in length wherein occurs, select intermediary characteristic value and minimum path; Described intermediary characteristic value and refer to each bar limit in each path the intermediary characteristic value and;

(A4): if obtain the K paths, then change step (A10) over to, otherwise variable k=1 is set, and change step (A5) over to;

(A5): sorted from small to large according to length in the path among the Ps, then from Ps, select k paths, p _k=(v ₁, v ₂..., v _i, v _j..., v _n), wherein, n is the number of nodes on this path, it is 0 integer that i, j are initial value;

(A6): to any 0＜j＜n, first with p _kLink (v _i, v _j) length is made as infinitely, obtain from v in network G afterwards _lTo v _nShortest path p _d, and with its adding set B, repeating step (A6) is until j=n-1;

(A7): i is added 1, and each linkage length is initial value among the recovery G, and repeating step (A6) is until i=n-2;

(A8): from set B, select intermediary's characteristic value and a minimum paths p ', and it is added set Ps;

(A9): if obtain the K paths, then change step (A10) over to, otherwise k is added 1, and repeating step (A5) until there is not any path among the B, then changes step (A10) to step (A9) over to;

(A10) finish.

Described described K bar based on each data stream transmitting request is the set P in overlapping transmission path not _i, iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision that maximizes minimum fair rule and is comprised:

(B1): based on the information of forecasting to bandwidth resources expense between the data central network, the network switch of utilizing time extension network switch method will have dynamic idle bandwidth resource is the passive flow network;

(B2): based on described passive flow network, the minimum fair many Commodity Flows linear programming model of maximization is set up in all data stream transmitting requests;

(B3): find the solution iteratively the minimum fair many Commodity Flows linear programming model of described maximization, draw maximum transmitted flow and the corresponding data transfer path of each data stream transmitting request.

The described time extension network switch method of described step (B1) is achieved in that Internet resources is extended from time dimension, and the dynamically bandwidth resource of the network with dynamic idle bandwidth resource and the storage resources ability unification of node are transformed on the passive flow network.

The minimum fair many Commodity Flows linear programming model of described maximization in the described step (B2) is:

maximize λ

$s.t.\underset{r_i\∈R}{\mathrm{\Σ}}\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≤C_e,\∀e\∈E,e\∈P_i---\left(1\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥\mathrm{\λ}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{unsat}}---\left(2\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥{\mathrm{\λ}}_{q_i}^{\mathrm{sat}}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{sat}}---\left(3\right)$

$f_p^{r_i}\≥0,\mathrm{\λ}\≥0,\∀p\∈P_{i=1...K}---\left(4\right)$

Wherein, r _iBe the request of a data flow transmission, R is the set of all data stream transmitting requests,

Be illustrated on the p of path and distribute to r _iBandwidth value, C _eRepresent the bandwidth resources of link e, E={e ₁, e ₂..., e _m, be the set of all link e in the network, λ is saturated allocated bandwidth ratio value, dem _iR _iThe data volume of transmission, R _UnsatBe the unsaturation request set,

For those of trying to achieve the maximum transmitted flow are asked corresponding λ value, R _SatBe saturated request set;

(1), (2), (3) and (4) these four formulas are constraints.

Described step (B3) comprising:

S1: set saturated request set R _SatBe empty, unsaturation request set R _UnsatComprise all data stream transmitting request r _i

S2: find the solution maximizing minimum fair many Commodity Flows linear programming model, namely satisfy at the same time under the prerequisite of (1), (2), (3) and (4) these four formulas, obtain maximum λ value;

S3: filter out the request set R that does not have unnecessary concatenation ability _Tmp: according to the situation of utilizing of the existing transmission path bandwidth resource of data stream transmitting request, if the bandwidth resources of certain data stream transmitting request are all taken, then it is selected into request set R _Tmp

S4: for R _TmpIn each the request r _i, with R _UnsatBe set as and only comprise a request r _i, R _SatThen be set as to comprise and remove request r among the request set R _iIn addition remaining whole requests, wherein original saturated request

Value remains unchanged, and the saturation value of unsaturation request then is set as λ, utilizes these values to upgrade the minimum fair many Commodity Flows linear programming model of maximization after obtaining upgrading after (2) formulas and (3) formula;

S5: find the solution the minimum fair many Commodity Flows linear programming model of maximization after the described renewal, obtain new λ _TmpIf λ _TmpEquate with λ, then ask r _iBe judged as really saturated, with r _iJoin R _Sat, and λ corresponding to record, if λ _TmpUnequal with λ, then change step S4 over to;

S6: repeatedly repeating step S2 is to step S5, until R _UnsatBe sky, obtain optimum λ;

S7: output maximum transmitted flow and corresponding data transfer path, described maximum transmitted flow refers to the λ of described optimum and the product of link bandwidth; Described data transfer path is included in corresponding with the λ of optimum

In.

Compared with prior art, the invention has the beneficial effects as follows: with document " M.Allalouf and Y.Shavitt; " Centralized and distributed algorithms for routing and weighted max-min fair bandwidth allocation; " Networking, IEEE/ACM Transactions on, vol.16, no.5, pp.1015-1024,2008 " and document " D.Nace, L.Nhat Doan, O.Klopfenstein, and A.Bashllari, " Max-min fairness in multi-commodity flows, " Computers ﹠amp; Operations Research, vol.35, no.2, pp.557-573, the MMF dispatching algorithm (OPT-MMF) of the optimum that 2008 " proposes is compared; the inventive method has solved under the prerequisite of more excellent allocated bandwidth value in assurance, greatly reduces its corresponding time overhead, has extremely strong practicality.

Description of drawings

Fig. 1 is the Path selection example of rule one.

Fig. 2 is the Path selection example of rule two.

Fig. 3 is Softlayer data center network topology schematic diagram.

Fig. 4-the 1st is the comparison diagram of Max-Min Fair bandwidth resource allocation total amount of three kinds of algorithms of 100 o'clock in the request scale.

Fig. 4-the 2nd is the comparison diagram of Max-Min Fair bandwidth resource allocation total amount of three kinds of algorithms of 500 o'clock in the request scale.

Fig. 4-the 3rd is the comparison diagram of Max-Min Fair bandwidth resource allocation total amount of three kinds of algorithms of 700 o'clock in the request scale.

Fig. 4-the 4th is the comparison diagram of Max-Min Fair bandwidth resource allocation total amount of three kinds of algorithms of 1000 o'clock in the request scale.

Fig. 5-the 1st is the comparison diagram of time overhead of three kinds of algorithms of 100 o'clock in the request scale.

Fig. 5-the 2nd is the comparison diagram of time overhead of three kinds of algorithms of 500 o'clock in the request scale.

Fig. 5-the 3rd is the comparison diagram of time overhead of three kinds of algorithms of 700 o'clock in the request scale.

Fig. 5-the 4th is the comparison diagram of time overhead of three kinds of algorithms of 1000 o'clock in the request scale.

Fig. 6 is the Efficient Solution scene of the minimum fair many Commodity Flows allocative decision of maximization in the embodiment of the invention.

Fig. 7 is the step block diagram that the inventive method maximizes minimum fair multiple data stream transmission dispatching method.

Embodiment

Below in conjunction with accompanying drawing the present invention is described in further detail:

As shown in Figure 7, the inventive method is based on following steps:

Input data center network topology information and data flow solicited message, the intermediary characteristic value on every limit in the computing network topology;

Step 1, the intermediary on limit in the topology Network Based (Edge Betweenness) characteristic value is assessed the different paths of point-to-point transmission, for the not set P in overlapping transmission path of specific K bar is selected in each data stream transmitting request _i

Step 2 is based on the K bar of each request set P in overlapping transmission path not _i, iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision that maximizes minimum fair rule;

Lower mask body makes an explanation to each step:

Step 1, the intermediary on limit in the topology Network Based (Edge Betweenness) characteristic value is assessed the different paths of point-to-point transmission, for not overlapping transmission set of paths of specific K bar is selected in each data flow request:

In order to improve the utilance of Internet resources, the present invention distributes many transmission paths for each request.Each data stream transmitting request the link circuit resource ability of how many transmission paths corresponding to it set of obtainable bandwidth resources closely related, so the selection of transmission paths process of data flow is most important.Draw through great many of experiments analysis and summary, when carrying out Path selection, in order to make each transmission acquisition request link bandwidth resource as much as possible, need to follow two rules: rule one, reduce the overlapping degree of each transmission path in the same request; Rule two, the overlapping degree of each transmission path between the different requests of reduction.Come this two rule is made an explanation below by example.

Rule one: as shown in Figure 1, given network G (V, E), wherein, V represents all node set in the network, and E represents the set of all links in the network, and the bandwidth capacity of link is shown in the Digital ID on each bar straight line among Fig. 1.There are source point and point of destination to be respectively v ₀And v ₂Data stream transmitting request r ₁Its article one transmission path and corresponding allocated bandwidth value are respectively p11=(v0, v1, v2) and fr1 (p11)=1 (because each request strives for obtaining allocated bandwidth resource as much as possible, so when request that transmission path p11 is distributed to during r1, the whole bandwidth resources on this path link, namely 1, request r1 so fr1 (p11)=1 have just all been given).When selecting the second transmission path for r1, two alternative path p12=(v0, v1, v4, v2) and p12 '=(v0, v3, v4, v2) arranged.Although the jumping figure of this two paths is identical, p12 ' is better than p12, because the overlapping degree of itself and p11 is lower, can guarantee that like this r1 obtains more bandwidth resources.

Rule two: as shown in Figure 2, data stream transmitting request r1 and r2 are arranged, its source point is respectively v0 and v3, and point of destination is respectively v2 and v8, and the allocated bandwidth value f (p) of its first three transmission paths p separately and correspondence is as shown in table 1.

Table 1

When selecting the 4th transmission paths for r1, two alternative paths are arranged, p14=(v0, v3, v5, v4, v2) and p14 '=(v0, v3, v1, v4, v2).Although the jumping figure of this two paths is identical, p14 ' is better than p14, because the overlapping degree of the path p21 of itself and r2 is lower, can guarantees like this itself and still less data transfer request competition, and obtain more bandwidth resources.

Based on above-mentioned two rules, the present invention has designed the routing algorithm based on the intermediary characteristic value on limit.Its basic thought is by calculating the intermediary value of each section link on every paths, assess the overlapping possibility in this paths and other paths, and then instructs the process of choosing in path.The intermediary characteristic value on network topology limit, that an one of attribute commonly used that be used for to estimate the significance level on limit in the graph theory (please refer to (1) L.Freeman, " A set of measures of centrality based on betweenness; " Sociometry, pp.35-41,1977 and (2) M.Girvan and M.Newman, " Community structure in social and biological networks; " Proceedings of the National Academy of Sciences, vol.99, no.12, p.7821,2002).It is defined as: Given Graph G (V, E), wherein the intermediary characteristic value of certain the limit e shortest path that equals any two points through the number of times of e and.According to above-mentioned definition, the intermediary characteristic value of a link is larger, means that so having more shortest paths comprises this link.Therefore, when selecting paths, except the jumping figure or bandwidth situation of weighing the path, also to consider the intermediary characteristic value in its each section path, preferentially select the little path of this value.

Based on above basic thought, described the concrete steps of the set in overlapping transmission path are not as follows for specific K bar is selected in each data flow request:

Step 1: given network G (V, E), source point and point of destination are respectively v _sAnd v _t, and the path number K that need to obtain;

Step 2: find the solution its shortest path p1 in network G, p1 is added set Ps, and (each data stream transmitting request has many transmission paths to realize the transmission of data flow, Ps is the set of paths of distributing to each transmission request), and the bandwidth resources that p1 is occupied deduct from G, obtain remaining network G r;

Step 3: (occurrence of K is artificial the setting to repeating step 2 on remaining network G r until number of path reaches K, K is larger, the allocation result of asking is near optimum, but computing cost is maximum, otherwise K is less, and allocation result becomes suboptimum, but computing cost is less) maybe can't obtain new shortest path and (namely on remaining network, do not have the path of any connection between transmitting terminal and the destination.), (many different transmission paths are arranged between transmitting terminal and the destination in network namely if the situation in the path of many equal in length wherein occurs, and the equal in length in path), select intermediary characteristic value and (be that the intermediary characteristic value addition on each bar limit in each path obtains and) minimum path;

Step 4: if obtain the K paths, then change step 10 over to, otherwise variable k=1 is set, and execution in step 5;

Step 5: (every paths is comprised of the multistage link according to length with the path among the Ps, the length in path refers to the number of links that it comprises herein, the number of links that comprises is fewer, the path is shorter) sort from small to large that (purpose of ordering is in order to select the short path of k, the new route length that generates based on modification k short path mode in the subsequent step so can be very not long), then from Ps, select k paths p _k=(v _l, v ₂..., v _i, v _j..., v _n), it is 0 integer that i wherein, j are initial value; Suppose that stretch directly comprises n node, each node v1, v2 ... vn represents, n is the number of nodes on this path, also is the label of last node in the path;

Step 6: to any 0＜j＜n, first with p _kLink (v _i, v _j) length is made as infinitely, obtain from v at G afterwards _lTo v _nShortest path p _d, and with its adding set B (this set is the set of an interim storage path candidate, and the new path that adds, back is selected from set B, until select the K paths), repeating step 6 is until j=n-1;

Step 7: i is added 1, and each linkage length is initial value among the recovery G, and repeating step 6 is until i=n-2;

Step 8: from set B, select intermediary's characteristic value and a minimum paths p ', and it is added set Ps;

Step 9: if obtain the K paths, then change step 10 over to, otherwise k is added 1, and repeating step 5 until there is not any path among the B, then changes step 10 to step 9 over to;

Step 10: finish.

The minimum fair many Commodity Flows linear programming model of maximization (after calculating candidate's transmission path for each transmission request, could begin to find the solution the allocated bandwidth value under the minimum fairness doctrine of maximization):

Utilize above-mentioned algorithm to find the solution corresponding data transfer path set P for each data stream transmitting request _i, use linear programming model to carry out modeling to maximizing the wide resource allocation problem of minimum fair many Commodity Flows.On given network G, supposing has N data transfer request, and each request r=(src, dest, dem) is a tlv triple, and wherein src represents that data send source point, and dest represents that data accept point of destination, and dem represents to ask the data volume transmitted.The available data transfer path set of each request is P _i, each request may have many different transmission paths, and the set that these paths form is exactly the path candidate set P of this request _i,

Be illustrated in and distribute to request r on the p of path _iBandwidth value, the minimum fair many Commodity Flows linear programming model of corresponding maximization (MMF linear programming model) is as follows:

maximize λ

$s.t.\underset{r_i\∈R}{\mathrm{\Σ}}\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≤C_e,\∀e\∈E,e\∈P_i---\left(1\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥\mathrm{\λ}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{unsat}}---\left(2\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥{\mathrm{\λ}}_{q_i}^{\mathrm{sat}}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{sat}}---\left(3\right)$

$f_p^{r_i}\≥0,\mathrm{\λ}\≥0,\∀p\∈P_{i=1...K}---\left(4\right)$

In the model, the s.t. meaning is the constraints that needs to satisfy (1), (2), (3) and (4) formula; The implication of block mold is that (1), (2), (3) and (4) these four formulas all are constraints, need to satisfy at the same time under the prerequisite of these four formulas, solve maximum λ value.

For those of trying to achieve the maximum transmitted flow are asked corresponding λ value;

In this model, request is divided into saturated request set R _SatWith unsaturation request set R _Unsat(those requests of maximum λ value have namely been obtained in saturated request, otherwise the request of also not obtaining maximum λ value is the unsaturation request.)。Saturated request refers under the MMF rule limits, can't continue to obtain the more request of the wide resource of multi-band; And the unsaturation request refers under the MMF rule limits, can further obtain the request of bandwidth resources.The target of finding the solution of model is the allocated bandwidth ratio value λ (ratio value that distributes bandwidth resources and bandwidth on demand resource is shown in (2) formula) of maximization unsaturation request; And to saturated request, only need then to guarantee that it is kept obtains existing allocated bandwidth ratio value

Get final product (shown in (3) formula); Simultaneously, the apportioning cost of bandwidth resources need to guarantee to be no more than the capacity limit (shown in (1) formula) of respective links;

Be real number with λ, free routing p is one (shown in (4) formula) in each request path set.

Step 2 is asked existing transmission path set based on each, and iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision (i.e. the fair many Commodity Flows allocated bandwidth derivation algorithm of maximization minimum) that maximizes minimum fair rule:

Transmission path set based on MMF linear programming model and each request, designed the not minimum fair multiple data stream dispatching algorithm (K-Disjoint Shortest Multi-Path Max-Min Fair, K-DSMP-MMF) of maximization of overlay path (below with method of the present invention referred to as K-DSMP-MMF) of efficient K bar.Its corresponding basic thought is: the allocated bandwidth resource that increases at first synchronously all unsaturation requests, until when certain request can't continue to obtain the wider resource of multi-band, this request is included in the middle of the set of saturated request, and the bandwidth money value of obtaining of record this request this moment.So repeat above-mentioned steps, until all requests are all saturated, just obtained optimum bandwidth allocation scheme.The algorithm concrete steps are described below:

Step 1: set saturated request set for empty, the unsaturation request set comprises whole transmission requests, solves whole transmission paths (being the not set in overlapping transmission path of described K bar) of each request according to the given network information;

Step 2: the MMF linear programming model is found the solution, draw allocated bandwidth ratio value λ;

Step 3: to request saturated checking the whether: at first filter out the request set R that does not have unnecessary concatenation ability _TmpFor R _TmpIn each the request r _i, with R _UnsatBe set as and only comprise request r _i, R _SatThen be set as to comprise to remain all and ask, wherein the saturation value of original saturated request remains unchanged, the existing bandwidth resource allocation ratio value of unsaturation request

Then be set as λ; Continue to find the solution the MMF linear programming model after the renewal, obtain new allocated bandwidth ratio value λ _TmpIf λ _TmpEquate with λ, then ask r _iCan be judged as really saturated, with r _iJoin R _Sat, and λ corresponding to record;

Step 4: repeating step 2 is to step 3, until R _UnsatBe sky.

Step 5: allocated bandwidth value and the corresponding data transfer path of the many transmission requests of output.

For the advantage of algorithm of the present invention is described better, with K-DSMP-MMF of the present invention and OPT-MMF and do not consider that the K-SMP-MMF dispatching algorithm of link plyability compares from " MMF bandwidth resource allocation total amount " and " Riming time of algorithm expense " two aspects.

Arranging of experiment is as follows: the True Data center backbone-network-mapping of choosing Softlayer company (please refer to Softlayer datacenter map, http://www.softlayer.com/advantages/network-overview/), comprise 11 data Centroids and 17 two-way backbone links, connected 200 host nodes under each data center's node.Data center's backbone bandwidth is set as 10GB, and each host node is according to the uplink and downlink bandwidth that 1Gb is arranged, and the data center network topology as shown in Figure 3.

Below by analyzing experimental data K-DSMP-MMF Algorithm Performance of the present invention is described.

When investigating " bandwidth resource allocation total amount ", successively ask quantity from 100-1000, be spaced apart 10 groups of experiments of 100, every group of experiment repeats 5 times and asks for its mean value.Wherein the sending and receiving of each request point is random generation, and its transmitted data amount is evenly distributed in the scope of 0.1TB-5TB.Compare with OPT-MMF and K-SMP-MMF algorithm, the three is 100 in the request scale, 500,700 and 1000 o'clock Max-Min Fair bandwidth resource allocation total amount such as Fig. 4-1 to shown in Fig. 4-4, can find: 1. by guaranteeing just that for each request distribution is no more than 18 paths it reaches optimal value, proof can be selected suitable set of paths by being thought of as each request, realizes reducing the purpose of bandwidth resource allocation complexity; 2.K-DSMP-MMF can use the maximization of finishing bandwidth resource allocation than K-SMP-MMF path still less, this is because it has considered the impact of the plyability factor in path when selecting paths.By the experiment of this group, can draw and select K-DSMP-MMF algorithm of the present invention, when being set, K=10 can obtain more excellent Max-Min-Fair apportioning cost.

When investigating " Riming time of algorithm expense " index, successively ask quantity from 100-1000, be spaced apart 10 groups of experiments of 100, every group of experiment repeats 5 times and asks for its mean value.Wherein the sending and receiving of each request point is random generation, and its transmitted data amount is evenly distributed in the scope of 0.1TB-5TB.Compare with OPT-MMF and K-SMP-MMF algorithm, the three is 100 in the request scale, 500,700 and 1000 o'clock Max-Min Fair find the solution Max-Min Fair bandwidth resource allocation result time overhead such as Fig. 5-1 to shown in Fig. 5-4, can find: 1. by reducing the path candidate set scale of each request, can significantly reduce the time scale expense of algorithm operation, when K=10, K-SMP-MMF and K-DSMP-MMF the two than lacking 63.7% the running time of OPT-MMF algorithm; 2.K-SMP-MMF and the time overhead of K-DSMP-MMF algorithm is linear growth with the increase of K, and the two is more or less the same.By these two groups experiments, can sum up and draw K-DSMP-MMF and can use seldom time, obtain more excellent Max-Min Fair Internet resources allocated bandwidth result.

The present invention can be applied to maximize in the Efficient Solution scene of minimum fair many Commodity Flows allocative decision.Fig. 6 has described the embodiment of a correspondence, and this network has 9 nodes, and the bandwidth resources situation of each link is shown in sign among Fig. 6.Two unit transmission requests are arranged, be respectively r ₁=(v ₀, v ₂, 1) and r ₂=(v ₃, v ₈, 1).Among this embodiment, as shown in table 2 through the allocative decision that the inventive method is found the solution when drawing K=4, allocated bandwidth value separately is respectively 3,2.Among this embodiment, to shown in Fig. 5-4, utilizing the used time of finding the solution of the inventive method is 12 seconds, and utilizes 22 seconds running times of OPT-MMF algorithm such as Fig. 5-1, and operation platform is Intel Core 2 Duo CPU 2GHz, the 2GB internal memory.Table 2 explanation can be by effectively selecting each transmission request candidate's transmission path (be each request selection 4 not overlay paths), try to achieve the MMF apportioning cost (namely 3,2) of suboptimum.And traditional method need to calculate whole candidate's transmission paths for each request when finding the solution the MMF apportioning cost, causes computing cost very large, and this can-4 given experimental datas find out from Fig. 5-1 to Fig. 5.

Table 2

Technique scheme is one embodiment of the present invention, for those skilled in the art, on the basis that the invention discloses application process and principle, be easy to make various types of improvement or distortion, and be not limited only to the described method of the above-mentioned embodiment of the present invention, therefore previously described mode is just preferred, and does not have restrictive meaning.

Claims (6)

1. one kind maximizes minimum fair multiple data stream transmission dispatching method, and it is characterized in that: described method comprises:

Input data center network topology information and data stream transmitting solicited message, the intermediary characteristic value on every limit in the computing network topology; Described data center network topology information comprises that the link connection between each data center concerns and the bandwidth capacity of every link; Described data stream transmitting solicited message comprises the data volume of transmitting terminal, destination and the request transmission of each data stream transmitting request;

Based on described intermediary characteristic value, the different paths of point-to-point transmission are assessed, for the not set P in overlapping transmission path of specific K bar is selected in each data stream transmitting request _iAnd

Based on the described K bar of each the data stream transmitting request set P in overlapping transmission path not _i, iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision that maximizes minimum fair rule.

2. the minimum fair multiple data stream transmission dispatching method of maximization according to claim 1, it is characterized in that: described based on described intermediary characteristic value, different paths to point-to-point transmission are assessed, for the not set P in overlapping transmission path of specific K bar is selected in each data stream transmitting request _iComprise:

(A1): given network G (V, E), source point and point of destination are respectively v _sAnd v _t, and the path number K that need to obtain;

(A2): find the solution its shortest path p1 in network G, p1 is added set Ps, and the bandwidth resources that p1 is occupied deduct from G, obtain remaining network G r; Described Ps is the set of paths of distributing to each data stream transmitting request;

(A3): repeating step on remaining network G r (A2) maybe can't be obtained new shortest path until number of path reaches K; If the situation in the path of many equal in length wherein occurs, select intermediary characteristic value and minimum path; Described intermediary characteristic value and refer to each bar limit in each path the intermediary characteristic value and;

(A4): if obtain the K paths, then change step (A10) over to, otherwise variable k=1 is set, and change step (A5) over to;

(A5): sorted from small to large according to length in the path among the Ps, then from Ps, select k paths, p _k=(v ₁, v ₂..., v _i, v _j..., v _n), wherein, n is the number of nodes on this path, it is 0 integer that i, j are initial value;

(A6): to any 0＜j＜n, first with p _kLink (v _i, v _j) length is made as infinitely, obtain from v in network G afterwards _lTo v _nShortest path p _d, and with its adding set B, repeating step (A6) is until j=n-1;

(A7): i is added 1, and each linkage length is initial value among the recovery G, and repeating step (A6) is until i=n-2;

(A8): from set B, select intermediary's characteristic value and a minimum paths p ', and it is added set Ps;

(A9): if obtain the K paths, then change step (A10) over to, otherwise k is added 1, and repeating step (A5) until there is not any path among the B, then changes step (A10) to step (A9) over to;

(A10) finish.

3. the minimum fair multiple data stream transmission dispatching method of maximization according to claim 2 is characterized in that: described described K bar based on each data stream transmitting request is the set P in overlapping transmission path not _i, iteration is obtained its corresponding optimum satisfied network bandwidth resources allocative decision that maximizes minimum fair rule and is comprised:

(B1): based on the information of forecasting to bandwidth resources expense between the data central network, the network switch of utilizing time extension network switch method will have dynamic idle bandwidth resource is the passive flow network;

(B2): based on described passive flow network, the minimum fair many Commodity Flows linear programming model of maximization is set up in all data stream transmitting requests;

(B3): find the solution iteratively the minimum fair many Commodity Flows linear programming model of described maximization, draw maximum transmitted flow and the corresponding data transfer path of each data stream transmitting request.

4. the minimum fair multiple data stream transmission dispatching method of maximization according to claim 3, it is characterized in that: the described time extension network switch method of described step (B1) is achieved in that Internet resources is extended from time dimension, and the dynamically bandwidth resource of the network with dynamic idle bandwidth resource and the storage resources ability unification of node are transformed on the passive flow network.

5. the minimum fair multiple data stream transmission dispatching method of maximization according to claim 4 is characterized in that: the minimum fair many Commodity Flows linear programming model of the described maximization in the described step (B2) is:

maximize λ

$s.t.\underset{r_i\∈R}{\mathrm{\Σ}}\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≤C_e,\∀e\∈E,e\∈P_i---\left(1\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥\mathrm{\λ}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{unsat}}---\left(2\right)$

$\underset{p\∈P_i}{\mathrm{\Σ}}{f}_p^{r_i}\≥{\mathrm{\λ}}_{q_i}^{\mathrm{sat}}\·{\mathrm{dem}}_i,\∀r_i\∈R_{\mathrm{sat}}---\left(3\right)$

$f_p^{r_i}\≥0,\mathrm{\λ}\≥0,\∀p\∈P_{i=1...K}---\left(4\right)$

Wherein, r _iBe the request of a data flow transmission, R is the set of all data stream transmitting requests,

Be illustrated on the p of path and distribute to r _iBandwidth value, C _eRepresent the bandwidth resources of link e, E={e ₁, e ₂..., e _m, be the set of all link e in the network, λ is saturated allocated bandwidth ratio value, dem _iR _iThe data volume of transmission, R _UnsatBe the unsaturation request set, For those of trying to achieve the maximum transmitted flow are asked corresponding λ value, R _SatBe saturated request set;

(1), (2), (3) and (4) these four formulas are constraints.

6. the minimum fair multiple data stream transmission dispatching method of maximization according to claim 5, it is characterized in that: described step (B3) comprising:

S1: set saturated request set R _SatBe empty, unsaturation request set R _UnsatComprise all data stream transmitting request r _i

S2: find the solution maximizing minimum fair many Commodity Flows linear programming model, namely satisfy at the same time under the prerequisite of (1), (2), (3) and (4) these four formulas, obtain maximum λ value;

S3: filter out the request set R that does not have unnecessary concatenation ability _Tmp: according to the situation of utilizing of the existing transmission path bandwidth resource of data stream transmitting request, if the bandwidth resources of certain data stream transmitting request are all taken, then it is selected into request set R _Tmp

S4: for R _TmpIn each the request r _i, with R _UnsatBe set as and only comprise a request r _i, R _SatThen be set as to comprise and remove request r among the request set R _iIn addition remaining whole requests, wherein original saturated request

Value remains unchanged, and the saturation value of unsaturation request then is set as λ, utilizes these values to upgrade the minimum fair many Commodity Flows linear programming model of maximization after obtaining upgrading after (2) formulas and (3) formula;

S5: find the solution the minimum fair many Commodity Flows linear programming model of maximization after the described renewal, obtain new λ _TmpIf λ _TmpEquate with λ, then ask r _iBe judged as really saturated, with r _iJoin R _Sat, and λ corresponding to record, if λ _TmpUnequal with λ, then change step S4 over to;

S6: repeatedly repeating step S2 is to step S5, until R _UnsatBe sky, obtain optimum λ;

S7: output maximum transmitted flow and corresponding data transfer path, described maximum transmitted flow refers to the λ of described optimum and the product of link bandwidth; Described data transfer path is included in corresponding with the λ of optimum In.

Images